Ivy Bridge question

[Meghan54]

I asked, so I care. If there is legit info that confirms which current mobos (if any) will supprt IB when it comes out, I would like to see the source.

Well, Asus states that all their current non-multi-gpu Socket 1155 motherboards will support PCIe 3.0 and all their S1155 mb's will support IB with the proper BIOS update.

Take it for what it's worth......

http://promos.asus.com/US/Z68Gen3/

(At bottom of linked page, small print just above the first chart....)

 

[toyota]

so Ivy Bridge 1155 will have basically the same IPC and still be just a quad for i5 and a quad with HT for i7? I do not like that because my 2500k can already hit almost 90% cpu usage during BF 3 with just a gtx570. and I do not care for HT as it can actually cause a 1-2 fps loss in performance in some cases. and even in games that can use more than 4 threads it does basically nothing.

I really wish they could have a 6 core cpu with no HT for the top of line 1155 Ivy Bridge but that is pretty unlikely. I just think getting yet another quad with same IPC next year seems pretty short sided for many people even if it does clock higher. maybe they can at least improve upon HT but that too is not very likely.

 

[mrjoltcola]

So you can just disable HT in the BIOS, or are you concerned that HT puts the CPU into a higher price category?

HT is probably as good as it'll get, next to full-core.

You shouldn't write HT off unless you really have benchmarked it. It really does work for threads that are somewhat memory intensive. You can't keep a core 100% pegged with a thread that has cache / memory stalls, and HT interleaves very well. There is a good example in Anand's Bulldozer benchmark, I think in one of the encryption benchmarks, where the Bulldozer actually jumps way ahead of the quads & 6x-Phenom, except for the 2600K which only comes out ahead due to HT. Did you actually see avg FPS over, say a 5-minute benchmark drop 1-2FPS? That is almost statistical noise, unless you tested in single player mode.

In any case, HT is a cheap option, and you can disable it right?

I agree, the 6-core needs updating, because a lot of people are forced to choose between Nehalem and Sandy Bridge right now, depending on their needs. The sheer superiority of Sandy Bridge architecture is what prompted me to update my old Conroe systems, so even though I wanted 6-core, I wasn't about to invest in Nehalem; but will gladly pay 50% more than my 2600K for a 6-core Sandy. So we'll get that in SB-E, and I assume IV-E will follow.

I won't complain either way, unless the chips are priced like the previous hexacore Extremes at over $1K. I think with the price bar has been lowered by 2500K/2600K, so I don't see them charging more than 500-600 for a hexacore SB-E.

 

[toyota]

I am only talking about HT in regards to gaming. nearly all of us here have decent quads so it would be nice to finally get more than 4 real cores from Intel at the same price bracket quads have been at for a while.

 

[mrjoltcola]

I am only talking about HT in regards to gaming. nearly all of us here have decent quads so it would be nice to finally get more than 4 real cores from Intel at the same price bracket quads have been at for a while.

I truly think we will. Now that the 2600K has set expectations at $300, I don't see how Intel could charge much more than $450-$500 for 6-core Sandy. Maybe I'm wrong. Though it sounds like you may be even asking for a 6-core non-HT for $300-ish, ala a 6-core 2500K big-brother? That would be nice, though I don't see it happening at that price for a while, since it would spank the 2600K way too badly, and now that Bulldozer is out, Intel can coast for a while and milk the market.

 

[StrangerGuy]

History probably repeats itself:

BD = Phenom I flop

Q6600 then = 2500K now.

IB = would be what 45nm C2Qs to Q6600. Slightly faster but not worth upgrading over.

 

[arredondo]

^^^ Heh, that makes me an efficient rig builder. My current machine is a Q6600. I'm in the process of upgrading to a i2500K machine. I'm sure I'll be happy with this until the new hotness comes along in five years or so.

Back on topic, I just ordered the ASRock Z68 Extreme3 Gen3 board for $115. Thanks for everyone's help and comments.

 

[greenhawk]

I just think getting yet another quad with same IPC next year seems pretty short sided for many people even if it does clock higher.

Just to show that upgrading at each step is not a smart move cost wise.

 

[greenhawk]

I don't see how Intel could charge much more than $450-$500 for 6-core Sandy. Maybe I'm wrong.

that is what would be considered a good price, but as with all new designs, intel will want the "new item" tax, so more like looking at $600-700 for that CPU, espically if intel want to make the sb-e's $500-600 chip look more attractive to those wanting 6 cores.

 

[arredondo]

Having the option to upgrade from a Sandy Bridge processor to an Ivy Bridge processor is a good thing. One does not have to get the upgrade day one of a new launch. It's great to have the choice available a few years after they've come out to delay a forced upgrade to the next non-compatible generation of processors.

 

[gevorg]

Having the option to upgrade from a Sandy Bridge processor to an Ivy Bridge processor is a good thing. One does not have to get the upgrade day one of a new launch. It's great to have the choice available a few years after they've come out to delay a forced upgrade to the next non-compatible generation of processors.

Yep! I will wait until first round of price cuts, but even without them, its pretty tempting for current Sandy Bridge owners since there is no need to upgrade mobo/RAM.

 

[MichaelD]

Only in the past few months did I feel the need to upgrade from a Q9550 (S775)/DDR2 setup to a "new" SandyBridge/DDR3 setup. Prices on i5/i7's have come down a lot and DDR3 has plummeted. Now was the time.

I ran a Q6600 for three years until it died and ran the Q9550 for the past 18 months. I had a good GPU (4870X2 and now a 470GTX) so the box ran my games just fine.

In our enthusiast world the actual NEED to upgrade really isn't too often. If your hardware dies or you get a bigger monitor, for example. But the WANT for the latest, shiny stuff is difficult to resist, I know.

If I currently had an i7/i5 rig I wouldn't even be thinking about Ivy Bridge. PCI-E 3.0? By the time peripheral manufs' get on board with it, PCI-E 3.1 will be out (HDMI anyone?). That ESTIMATED 18% speed boost IB will have over SB remains to be seen and with SB chips hitting 4.5GHz+ on air, how much more do you really need?

If you're gaming, it's mostly all about the GPU anyway.

 

[IntelUser2000]

There should be an IPC gain. It's not big, and 20% promise includes the performance per clock gains.

However, there are important things like allowing the resources to be used more efficiently when Hyperthreading is enabled, but isn't used. It'll help in a corner case situation that is bound by the resource.

Should be another 3-5% gain that would be focused on the media side, and Hyperthreading changes.

You can't exactly plot out estimations based on the transistor performance graph. It'll be off few %, but for the sake of the conversation, the 18% increase for 22nm referred to 18% delay reduction, which turns out to be 22% gain.

 

[BlahBlahYouToo]

The IGP is also useful for productivity (second monitor). Hooking up two monitors on one dedicated GPU stops it from downclocking to idle/2d clocks, which increases power consumption/heat/fan noise

can i run 3 monitors with a dedicated card that has 2 DVI (gtx460) and the DVI on the Z68 board?

 

[StrangerGuy]

There should be an IPC gain. It's not big, and 20% promise includes the performance per clock gains.

However, there are important things like allowing the resources to be used more efficiently when Hyperthreading is enabled, but isn't used. It'll help in a corner case situation that is bound by the resource.

Should be another 3-5% gain that would be focused on the media side, and Hyperthreading changes.

You can't exactly plot out estimations based on the transistor performance graph. It'll be off few %, but for the sake of the conversation, the 18% increase for 22nm referred to 18% delay reduction, which turns out to be 22% gain.

http://www.anandtech.com/show/4830/intels-ivy-bridge-architecture-exposed/4

Anand's IB article says 4 to 6% IPC increase, HT performance penalty is reduced (I bet that got eliminated) and 75-80% voltage required for the same clocks. How Intel managed to even squeeze another ~5% IPC increase out of SB is already darn impressive to me.

 

[IntelUser2000]

http://www.anandtech.com/show/4830/intels-ivy-bridge-architecture-exposed/4

Anand's IB article says 4 to 6% IPC increase, HT performance penalty is reduced (I bet that got eliminated) and 75-80% voltage required for the same clocks. How Intel managed to even squeeze another ~5% IPC increase out of SB is already darn impressive to me.

Well, I'm just guessing, so is Anand, so 1-2% doesn't make a difference in reality. I'm estimating it won't be large as say Penryn had because it doesn't have cache changes. The gains might be somewhat less general than those chips.

 

[Raqia]

~20W lower power draw: http://vr-zone.com/articles/ivy-bri...d-forwards-compatibility-explained/13754.html